Página 18 dos resultados de 2566 itens digitais encontrados em 0.040 segundos

Temporal pattern recognition in multiparameter ICU data

Saeed, Mohammed
Fonte: Massachusetts Institute of Technology Publicador: Massachusetts Institute of Technology
Tipo: Tese de Doutorado Formato: 219 leaves
Português
Relevância na Pesquisa
37.265425%
Intensive Care Unit (ICU) patients are physiologically fragile and require vigilant monitoring and support. The myriad of data gathered from biosensors and clinical information systems has created a challenge for clinicians to assimilate and interpret such large volumes of data. Physiologic measurements in the ICU are inherently noisy, multidimensional, and can readily fluctuate in response to therapeutic interventions as well as evolving pathophysiologic states. ICU patient monitoring systems may potentially improve the efficiency, accuracy and timeliness of clinical decision-making in intensive care. However, the aforementioned characteristics of ICU data can pose a significant signal processing and pattern recognition challenge---often leading to false and clinically irrelevant alarms. We have developed a temporal database of several thousand ICU patient records to facilitate research in advanced monitoring systems. The MIMIC-II database includes high-resolution physiologic waveforms such as ECG, blood pressures waveforms, vital sign trends, laboratory data, fluid balance, therapy profiles, and clinical progress notes over each patient's ICU stay. We quantitatively and qualitatively characterize the MIMIC-II database and include examples of clinical studies that can be supported by its unique attributes. We also introduce a novel algorithm for identifying "similar" temporal patterns that may illuminate hidden information in physiologic time series. The discovery of multi-parameter temporal patterns that are predictive of physiologic instability may aid clinicians in optimizing care. In this thesis...

Dynamic load-balancing of StreamIt cluster computations; Dynamic load-balancing of Stream It cluster computations

Fellheimer, Eric Todd
Fonte: Massachusetts Institute of Technology Publicador: Massachusetts Institute of Technology
Tipo: Tese de Doutorado Formato: 147 p.
Português
Relevância na Pesquisa
37.265425%
This thesis discusses the design and implementation of a dynamic load-balancing mechanism for computationally distributed programs running on a cluster written in the StreamIt programming language. StreamIt is useful for streaming data applications such as MPEG codecs. The structure of the language carries a lot of static information, such as data rates and computational hierarchy, and therefore lends itself well to parallelization. This work details a simulator for StreamIt cluster computations used to measure metrics such as throughput. Built on top of this simulation is an agent-based market used for load balancing the computation at StreamIt check-points to adapt to exogenously changing loads on the nodes of the cluster. The market models the structure of the computation as a supply chain. Our experiments study the throughput produced by the market compared to other policies, as well as qualitative features such as stability.; by Eric Todd Fellheimer.; Thesis (M. Eng.)--Massachusetts Institute of Technology, Dept. of Electrical Engineering and Computer Science, 2006.; Includes bibliographical references (p. 145-147).

Instructor authoring tool : a step towards promoting dynamic lecture-style classrooms

Chen, Jessie I
Fonte: Massachusetts Institute of Technology Publicador: Massachusetts Institute of Technology
Tipo: Tese de Doutorado Formato: 64 leaves
Português
Relevância na Pesquisa
37.265425%
The Instructor Authoring Tool of the Classroom Learning Partner project provides an efficient tool for university professors and other teachers of lecture-style classrooms to construct lecture slides that can easily accommodate effective in-class exercises. In designing such an authoring tool, five criterions were taken into consideration: An enhanced ability to support instructors' use of in-class exercises by facilitating the slide-generation process, the use of Microsoft PowerPoint as a basic tool, enabling real-time feedback of student performance on exercises, the leveraging of past student mistakes and misunderstood concepts in enhancing teaching, and the support of an automatic tutoring system to be implemented at a later time. These criterions were successfully met by the resulting instructor authoring tool, though improvements towards facilitating the slide-generation process and support for real-time feedback of student performance have yet to be tested in a formal academic setting.; by Jessie I. Chen.; Thesis (M. Eng. and S.B.)--Massachusetts Institute of Technology, Dept. of Electrical Engineering and Computer Science, 2006.; Includes bibliographical references (leaves 59-64).

Variable block length coding for channels with feedback and cost constraints

Nakiboḡlu, Bariş
Fonte: Massachusetts Institute of Technology Publicador: Massachusetts Institute of Technology
Tipo: Tese de Doutorado Formato: 96 p.; 472133 bytes; 467434 bytes; application/pdf; application/pdf
Português
Relevância na Pesquisa
37.265425%
Variable-decoding-time/generalized block-coding schemes are investigated for discrete memoryless channels (DMC) with perfect feedback (error free, delay free, infinite capacity) under cost constraints. For a given number of messages and average error probability, upper and lower bounds are found for expected decoding time. These coincide with each other up to a proportionality constant which approaches one in a certain asymptotic sense. A resulting reliability function is found for variable decoding time DMC's with perfect feedback under a cost constraint. The results in this work generalize Burnashev's results, to the cost constrained case.; by Bariş Nakiboḡlu.; Thesis (S.M.)--Massachusetts Institute of Technology, Dept. of Electrical Engineering and Computer Science, 2005.; This electronic version was submitted by the student author. The certified thesis is available in the Institute Archives and Special Collections.; Includes bibliographical references (p. 95-96).

Design and fabrication of an electrically-activated photonic crystal nanocavity laser

Mattson, Eric (Eric Michael)
Fonte: Massachusetts Institute of Technology Publicador: Massachusetts Institute of Technology
Tipo: Tese de Doutorado Formato: 71 p.; 3534764 bytes; 3537697 bytes; application/pdf; application/pdf
Português
Relevância na Pesquisa
37.265425%
In the future, optical networks may see an expanded role not only in telecommunications, but also in computers and other common electronic devices. These optical networks will require small, on-chip light sources. By using the photonic crystal's ability to strongly confine light, photonic crystal lasers can be built very small and very efficient, making them ideal for photonic integrated circuits. This thesis describes the design and fabrication of an electrically-activated photonic crystal nanocavity laser using an active layer with quantum dots. Hydrogen silsesquioxane (HSQ) was studied as an electron-beam lithography resist, and reactive ion etching of AlGaAs and InGaAlP was investigated. The laser described herein is very small, only - 5 gm in length and width. The design is also very flexible. By simply changing the active material and the size and spacing of the holes which create the one-dimensional photonic crystals, the emission wavelength can be easily varied. The laser is anticipated to be more efficient than the current technology from both the energy and chip design standpoints, and should represent a major improvement in on-chip light sources.; Eric Mattson.; Thesis (S.M.)--Massachusetts Institute of Technology, Dept. of Electrical Engineering and Computer Science...

Eclat : automatic generation and classification of test inputs; Automatic generation and classification of test inputs

Pacheco, Carlos, S.M. Massachusetts Institute of Technology
Fonte: Massachusetts Institute of Technology Publicador: Massachusetts Institute of Technology
Tipo: Tese de Doutorado Formato: 54 p.; 2353622 bytes; 2355771 bytes; application/pdf; application/pdf
Português
Relevância na Pesquisa
37.265425%
This thesis describes a technique that selects, from a large set of test inputs, a small subset likely to reveal faults in the software under test. The technique takes a program or software component, plus a set of correct executions-say, from observations of the software running properly, or from an existing test suite that a user wishes to enhance. The technique first infers an operational model of the software's operation. Then, inputs whose operational pattern of execution differs from the model in specific ways are suggestive of faults. These inputs are further reduced by selecting only one input per operational pattern. The result is a small portion of the original inputs, deemed by the technique as most likely to reveal faults. Thus, the technique can also be seen as an error-detection technique. The thesis describes two additional techniques that complement test input selection. One is a technique for automatically producing an oracle (a set of assertions) for a test input from the operational model, thus transforming the test input into a test case. The other is a classification-guided test input generation technique that also makes use of operational models and patterns. When generating inputs, it filters out code sequences that are unlikely to contribute to legal inputs...

Energy scalability of on-chip interconnection networks; Energy scalability of OCN

Konstantakopoulos, Theodoros K., 1977-
Fonte: Massachusetts Institute of Technology Publicador: Massachusetts Institute of Technology
Tipo: Tese de Doutorado Formato: 198 p.
Português
Relevância na Pesquisa
37.265425%
On-chip interconnection networks (OCN) such as point-to-point networks and buses form the communication backbone in multiprocessor systems-on-a-chip, multicore processors, and tiled processors. OCNs consume significant portions of a chip's energy budget, so their energy analysis early in the design cycle becomes important for architectural design decisions. Although innumerable studies have examined OCN implementation and performance, there have been few energy analysis studies. This thesis develops an analytical framework for energy estimation in OCNs, for any given topology and arbitrary communication patterns, and presents OCN energy results based on both analytical communication models and real network traces from applications running on a tiled multicore processor. This thesis is the first work to address communication locality in analyzing multicore interconnect energy and to use real multicore interconnect traces extensively. The thesis compares the energy performance of point-to-point networks with buses for varying degrees of communication locality. The model accounts for wire length, switch energy, and network contention. This work is the first to examine network contention from the energy standpoint.; (cont.) The thesis presents a detailed analysis of the energy costs of a switch and shows that the estimated values for channel energy...

Efficient model learning for dialog management

Doshi, Finale (Finale P.)
Fonte: Massachusetts Institute of Technology Publicador: Massachusetts Institute of Technology
Tipo: Tese de Doutorado Formato: 122 p.
Português
Relevância na Pesquisa
37.265425%
Partially Observable Markov Decision Processes (POMDPs) have succeeded in many planning domains because they can optimally trade between actions that will increase an agent's knowledge about its environment and actions that will increase an agent's reward. However, POMDPs are defined with a large number of parameters which are difficult to specify from domain knowledge, and gathering enough data to specify the parameters a priori may be expensive. This work develops several efficient algorithms for learning the POMDP parameters online and demonstrates them on dialog manager for a robotic wheelchair. In particular, we show how a combination of specialized queries ("meta-actions") can enable us to create a robust dialog manager that avoids the pitfalls in other POMDP-learning approaches. The dialog manager's ability to reason about its uncertainty -- and take advantage of low-risk opportunities to reduce that uncertainty -- leads to more robust policy learning.; by Final Doshi.; Thesis (S.M.)--Massachusetts Institute of Technology, Dept. of Electrical Engineering and Computer Science, 2007.; This electronic version was submitted by the student author. The certified thesis is available in the Institute Archives and Special Collections.; Includes bibliographical references (p. 118-122).

Diabetes Research Vortex : a novel information management system for Type I diabetes; DRV : a novel information management system for T1D

Javanmardian, Kia
Fonte: Massachusetts Institute of Technology Publicador: Massachusetts Institute of Technology
Tipo: Tese de Doutorado Formato: 69 leaves; 3429612 bytes; 3432038 bytes; application/pdf; application/pdf
Português
Relevância na Pesquisa
37.265425%
Information management is becoming a necessary task for modem research laboratories. As scientific research within a particular domain progresses, the amount of data and publications in the domain increases drastically. Type one diabetes (TID) research is no exception. The DRV (Diabetes Research Vortex) addresses the information management challenge. The DRV's main objective is to provide a web accessible information management system for diabetes related publications and experimental data. The DRV was inspired by current T1D research paradigms. A user can choose to upload a publication or experimental data under one of four research categories: a strain, gene, Idd (Insulin Dependent Diabetic) genetic region, or custom keyword. The effectiveness of the DRV lies in the biological concepts it encapsulates. The DRV allows a user to view the system's contents by browsing a strain, gene, Idd, or custom keyword. Furthermore, the DRV allows a user to seamlessly browse among related strains, genes, and Idd's. The DRV links information between each of these domains and allows a user to intuitively acquire knowledge about a specific interest. The system can also be asked to mine publications from PubMed based the terms stored in the -four aforementioned categories. Furthermore...

Flexible application driven network striping over Wireless Wide Area Networks

Qureshi, Asfandyar
Fonte: Massachusetts Institute of Technology Publicador: Massachusetts Institute of Technology
Tipo: Tese de Doutorado Formato: 161 p.; 10306058 bytes; 10313618 bytes; application/pdf; application/pdf
Português
Relevância na Pesquisa
37.265425%
Inverse multiplexing, or network striping, allows the construction of a high-bandwidth virtual channel from a collection of multiple low-bandwidth network channels. Striping systems usually employ a packet scheduling policy that allows applications to be oblivious of the way in which packets are routed to specific network channels. Though this is appropriate for many applications, many other applications can benefit from an approach that explicitly involves the application in the determination of the striping policy. Horde is middleware that facilitates flexible striping over Wireless Wide Area Network (WWAN) channels. Horde is unusual in that it separates the striping policy from the striping mechanism. It allows applications to describe network Quality-of-Service (QoS) objectives that the striping mechanism attempts to satisfy. Horde can be used by a set of data streams, each with its own QoS policy, to stripe data over a set of WWAN channels. The WWAN QoS variations observed across different channels and in time, provide opportunities to modulate stream QoS through scheduling. The key technical challenge in Horde is giving applications control over certain aspects of the data striping operation while at the same time shielding the application from low-level details. Horde exports a set of flexible abstractions replacing the application's network stack. Horde allows applications to express their policy goals as succinct network-QoS objectives. Each objective says something...

Non-uniform emission studies of a magnetron injection gun

Marchewka, Chad D. (Chad Daniel)
Fonte: Massachusetts Institute of Technology Publicador: Massachusetts Institute of Technology
Tipo: Tese de Doutorado Formato: 139 leaves; 5927530 bytes; 6286714 bytes; application/pdf; application/pdf
Português
Relevância na Pesquisa
37.265425%
This thesis investigates the experimental measurement and theoretical simulation of the effects of azimuthal emission non-uniformity of a 96 kV, 40 amp magnetron injection gun (MIG) used in a gyrotron. The accomplishments of this thesis include: Experimental measurement of the azimuthal emission non-uniformity of the MIG gun; Simulation of the beam quality of the MIG gun using MICHELLE 3-D, the first simulation of a MIG electron beam with azimuthal non-uniformity; Benchmarking the MICHELLE 3-D code to other established gun optics codes; Evaluation of the effects on the velocity spread and pitch factor of azimuthal non-uniformity in the MIG gun, showing that the direct effect on the beam quality is very small; Design, fabrication, and testing in the gyrotron of a capacitive probe system divided into four quadrants to measure azimuthal asymmetries of the electron beam; Use of the capacitive probes to measure low-frequency (100 - 160 MHz) oscillations on the beam, the first measurement of such oscillations in a microsecond pulse length gyrotron; First results on testing a new MIG cathode for emission non-uniformity using a special test chamber built by Calabazas Creek Research.; (cont.) This research will contribute to our understanding of the properties of intense electron beams produced by MIG guns in high-power gyrotrons. MIG's are widely used in gyrotron oscillators and amplifiers for fusion applications to create a beam of gyrating electrons generally operating in the temperature limited regime of emission. Due to this dependence on the temperature of the cathode...

Free text phrase encoding and information extraction from medical notes

Shu, Jennifer (Jennifer J.)
Fonte: Massachusetts Institute of Technology Publicador: Massachusetts Institute of Technology
Tipo: Tese de Doutorado Formato: 90 p.
Português
Relevância na Pesquisa
37.265425%
The Laboratory for Computational Physiology is collecting a large database of patient signals and clinical data from critically ill patients in hospital intensive care units (ICUs). The data will be used as a research resource to support the development of an advanced patient monitoring system for ICUs. Important pathophysiologic events in the patient data streams must be recognized and annotated by expert clinicians in order to create a "gold standard" database for training and evaluating automated monitoring systems. Annotating the database requires, among other things, analyzing and extracting important clinical information from textual patient data such as nursing admission and progress notes, and using the data to define and document important clinical events during the patient's ICU stay. Two major text-related annotation issues are addressed in this research. First, the documented clinical events must be described in a standardized vocabulary suitable for machine analysis. Second, an advanced monitoring system would need an automated way to extract meaning from the nursing notes, as part of its decision-making process. The thesis presents and evaluates methods to code significant clinical events into standardized terminology and to automatically extract significant information from free-text medical notes.; by Jennifer Shu.; Thesis (M. Eng.)--Massachusetts Institute of Technology...

A framework for cooperative localization in ultra-wideband wireless networks

Lien, Jaime
Fonte: Massachusetts Institute of Technology Publicador: Massachusetts Institute of Technology
Tipo: Tese de Doutorado Formato: 93 p.
Português
Relevância na Pesquisa
37.265425%
Location-aware technologies have the potential to revolutionize computing, cellular services, sensor networks, and many other commercial, military, and social applications. In wireless networks, accurate information about an agent's location can give meaning to observed data and facilitate the agent's interactions with its surroundings and neighbors. Determining the location of one or more agents, known as localization or positioning, is a fundamental challenge. Most existing localization methods rely on existing infrastructure and hence lack the flexibility and robustness necessary for large ad-hoc networks. In this thesis, we describe a framework for localization that overcomes these limitations by utilizing cooperation: the agents in the network work together to determine their individual locations. We derive a practical algorithm for cooperative localization by formulating the problem as a factor graph and applying the sum-product algorithm. Each agent uses relative positioning measurements and probabilistic location information from its neighbors to iteratively update its location estimate. We investigate the performance of this algorithm in a network of ultra-wideband (UWB) nodes, which are well-suited for localization due to their potential to measure inter-node distances with high accuracy. Realistic models of UWB ranging error...

Deniable Ring Signatures

Reich, Eitan
Fonte: Massachusetts Institute of Technology Publicador: Massachusetts Institute of Technology
Tipo: Tese de Doutorado Formato: 57 p.
Português
Relevância na Pesquisa
37.265425%
Ring Signatures were developed by Rivest, Shamir and Tauman, in a paper titled How to Leak a Secret, as a cryptographically secure way to authenticate messages with respect to ad-hoc groups while still maintaining the signer's anonymity. While their initial scheme assumed the existence of random oracles, in 2005 a scheme was developed that does not use random oracles and meets the strongest security definitions known in the literature. We argue that this scheme is not deniable, meaning if someone signs a message with respect to a ring of possible signers, and at a later time the secret keys of all of the possible signers are confiscated (including the author), then the author's anonymity is no longer guaranteed. We propose a modification to the scheme that guarantees anonymity even in this situation, using a scheme that depends on ring signature users generating keys that do not distinguish them from other users who did not intend to participate in ring signature schemes, so that our scheme can truly be called a deniable ring signature scheme.; by Eitan Reich.; Thesis (M. Eng.)--Massachusetts Institute of Technology, Dept. of Electrical Engineering and Computer Science, 2007.; Includes bibliographical references (p. 55-57).

Computational soundness for standard assumptions of formal cryptography

Herzog, Jonathan, 1975-
Fonte: Massachusetts Institute of Technology Publicador: Massachusetts Institute of Technology
Tipo: Tese de Doutorado Formato: 100 p.; 4682230 bytes; 4693726 bytes; application/pdf; application/pdf
Português
Relevância na Pesquisa
37.265425%
This implementation is conceptually simple, and relies only on general assumptions. Specifically, it can be thought of as a 'self-referential' variation on a well-known encryption scheme. 4. Lastly, we show how the computational soundness of the Dolev-Yao model can be maintained even as it is extended to include new operators. In particular, we show how the Diffie-Hellman key-agreement scheme and the computational Diffie-Hellman assumption can be added to the Dolev-Yao model in a computationally sound way.; The Dolev-Yao model is a useful and well-known framework in which to analyze security protocols. However, it models the messages of the protocol at a very high level and makes extremely strong assumptions about the power of the adversary. The computational model of cryptography, on the other hand, takes a much lower-level view of messages and uses much weaker assumptions. Despite the large differences between these two models, we have been able to show that there exists a relationship between them. Previous results of ours demonstrate that certain kinds of computational cryptography can result in an equivalence of sorts between the formal and computational adversary. Specifically: * We gave an interpretation to the messages of the Dolev-Yao model in terms of computational cryptography...

Translingual grammar induction for conversational systems

Lee, John Sie Yuen, 1977-
Fonte: Massachusetts Institute of Technology Publicador: Massachusetts Institute of Technology
Tipo: Tese de Doutorado Formato: 77 p.; 2169365 bytes; 2177448 bytes; application/pdf; application/pdf
Português
Relevância na Pesquisa
37.265425%
We propose an induction algorithm to semi-automate grammar authoring in an interlingua-based machine translation framework. This algorithm is designed for restricted domains within the context of multilingual conversational systems. It uses a pre-existing one-way translation system from some other language to the target language as prior information. It then infers a grammar for the target language. We demonstrate the system's effectiveness on a weather domain and on a travel domain. We automatically induced Chinese and French grammars for these domains from their English counterparts, and then showed that they can produce high-quality interlingua to be used in translation.; by John Sie Yuen Lee.; Thesis (S.M.)--Massachusetts Institute of Technology, Dept. of Electrical Engineering and Computer Science, 2004.; Includes bibliographical references (p. 75-77).

Rick arbitrage : analysis and trading systems

Naheta, Akshay, 1981-
Fonte: Massachusetts Institute of Technology Publicador: Massachusetts Institute of Technology
Tipo: Tese de Doutorado Formato: 60 leaves; 2258460 bytes; 2263896 bytes; application/pdf; application/pdf
Português
Relevância na Pesquisa
37.265425%
In this thesis we quantify the risk arbitrage investment process and create trading strategies that generate positive risk-adjusted returns. We use a sample of 895 stock swap mergers, cash mergers, and cash tender offers during 1998-2004Q2. We test the market efficiency hypothesis, and after accounting for transaction costs, we find that our risk arbitrage strategies generate annual risk-adjusted returns in excess of 4.5%. The research also obtains various other merger statistics, and relates them to a variety of economic indicators and merger timing models, as described in past work. We also estimate conditional probabilities of a merger's success, using a deal characteristic-driven prediction model, and combine it with market-implied probabilities. Our analysis suggests that the probability of success of a merger depends on a deal's characteristics. Further, it implies that one can improve on the market-implied estimates thereby creating trading opportunities. The analytical results achieved in this thesis can be used as the foundation for building an effective risk arbitrage trading platform.; by Akshay Naheta.; Thesis (S.M.)--Massachusetts Institute of Technology, Dept. of Electrical Engineering and Computer Science, 2004.; Includes bibliographical references (leaves 59-60).

User-directed sketch interpretation

Notowidigdo, Matthew J. (Matthew Joseph), 1981-
Fonte: Massachusetts Institute of Technology Publicador: Massachusetts Institute of Technology
Tipo: Tese de Doutorado Formato: 92 p.; 4083296 bytes; 4093811 bytes; application/pdf; application/pdf
Português
Relevância na Pesquisa
37.265425%
I present a novel approach to creating structured diagrams (such as flow charts and object diagrams) by combining an off-line sketch recognition system with the user interface of a traditional structured graphics editor. The system, called UDSI (user-directed sketch interpretation), aims to provide drawing freedom by allowing the user to sketch entirely off-line using a pure pen-and-paper interface. The results of the drawing can then be presented to UDSI, which recognizes shapes and lines and text areas that the user can then polish as desired. The system can infer multiple interpretations for a given sketch, to aid during the user's polishing stage. The UDSI program offers three novel features. First, it implements a greedy algorithm for determing alternative interpretations of the user's original pen drawing. Second, it introduces a user interface for selecting from these multiple candidate interpretations. Third, it implements a circle recognizer using a novel circle-detection algorithm and combines it with other hand-coded recognizers to provide a robust sketch recognition system.; by Matthew J. Notowidigdo.; Thesis (M. Eng.)--Massachusetts Institute of Technology, Dept. of Electrical Engineering and Computer Science, 2004.; Includes bibliographical references (p. 91-92).

Cyclic exchange neighborhood search technique for the K-means clustering problem

Thirathon, Nattavude, 1980-
Fonte: Massachusetts Institute of Technology Publicador: Massachusetts Institute of Technology
Tipo: Tese de Doutorado Formato: 152 p.; 6387181 bytes; 6407754 bytes; application/pdf; application/pdf
Português
Relevância na Pesquisa
37.265425%
Cyclic Exchange is an application of the cyclic transfers neighborhood search technique for the k-means clustering problem. Neighbors of a feasible solution are obtained by moving points between clusters in a cycle. This method attempts to improve local minima obtained by the well-known Lloyd's algorithm. Although the results did not establish usefulness of Cyclic Exchange, our experiments reveal some insights on the k-means clustering and Lloyd's algorithm. While Lloyd's algorithm finds the best local optimum within a thousand iterations for most datasets, it repeatedly finds better local minima after several thousand iterations for some other datasets. For the latter case, Cyclic Exchange also finds better solutions than Lloyd's algorihtm. Although we are unable to identify the features that lead Cyclic Exchange to perform better, our results verify the robustness of Lloyd's algorithm in most datasets.; by Nattavude Thirathon.; Thesis (M. Eng.)--Massachusetts Institute of Technology, Dept. of Electrical Engineering and Computer Science, 2004.; Includes bibliographical references (p. 151-152).

Learning commonsense categorical knowledge in a thread memory system

Stamatoiu, Oana L. (Oana Liana), 1981-
Fonte: Massachusetts Institute of Technology Publicador: Massachusetts Institute of Technology
Tipo: Tese de Doutorado Formato: 92 p.; 5162055 bytes; 5172554 bytes; application/pdf; application/pdf
Português
Relevância na Pesquisa
37.265425%
If we are to understand how we can build machines capable of broad purpose learning and reasoning, we must first aim to build systems that can represent, acquire, and reason about the kinds of commonsense knowledge that we humans have about the world. This endeavor suggests steps such as identifying the kinds of knowledge people commonly have about the world, constructing suitable knowledge representations, and exploring the mechanisms that people use to make judgments about the everyday world. In this work, I contribute to these goals by proposing an architecture for a system that can learn commonsense knowledge about the properties and behavior of objects in the world. The architecture described here augments previous machine learning systems in four ways: (1) it relies on a seven dimensional notion of context, built from information recently given to the system, to learn and reason about objects' properties; (2) it has multiple methods that it can use to reason about objects, so that when one method fails, it can fall back on others; (3) it illustrates the usefulness of reasoning about objects by thinking about their similarity to other, better known objects, and by inferring properties of objects from the categories that they belong to; and (4) it represents an attempt to build a autonomous learner and reasoner...